We use fly color-vision circuitry as a model to study how neurons form complex connections or circuits during development. The fly retina contains three types of photoreceptors, R1-6, R7, and R8, each of which responds to a specific spectrum of light and connects to a specific layer in the brain in the retinotopic fashion. We are focusing on the UV-sensitive R7 neurons which connect to the M6 layer in the medulla neuropil in the retinotopic fashion. Using wavelength-selection behaviors, we have isolated a number of mutants in which various aspects of R7-brain connectivity are affected. We are currently cloning and analyzing these mutants. [unreadable] [unreadable] Our developmental analyses revealed that the layer-specific connections mediated by R7s develop in two distinct stages. During the late larval and early pupal stage (the first target-selection stage), R7 neurons sequentially differentiate and project axons into the R7-temporary layer where they remain for 1-2 days. During the late pupal stage (the second target-selection stage), all R7 growth cones regain motility and synchronously project into the destined layer, the M6 layer. Cell-ablation analysis revealed that at the first stage, layer-specific connectivity is dictated by the interactions between R7 afferents and the dendrites of medulla neurons. The characterization of the development of R7-brain connections provides a framework to study the isolated mutants.[unreadable] [unreadable] In a forward genetic screen based wavelength-selection behavior, we isolated N-cadherin, LAR, and a novel mutant ovs (overshoot), all of which affect the development of R7 layer-specific connections in a cell-autonomous fashion. N-cadherin encodes a homophilic calcium-dependent adhesion molecule and LAR encodes a receptor tyrosine phosphatase. N-cadherin or LAR mutant R7 axons retract to the superficial R8-recipient layer during development. Mosaic analyses showed that N-cadherin is required in both R7 afferents and medulla neurons to direct R7-specific connections, suggesting that N-cadherin mediates homophilic interactions between R7 growth cones and dendrites of medulla neurons. Structure-function analyses further revealed that the cytoplasmic domain, hence the catenin-binding activity, of N-cadherin is largely dispensable for R7 layer-specific targeting. These results indicate that the adhesive, but not the signaling, activity of N-cadherin is essential for the development of R7-layer-specific connections. In contrast to N-cadherin, ovs mutant R7 axons overshoot the M6 layer and terminate at a deeper layer. We hypothesize that ovs provides the opposing (or balancing) effect on the R7 growth cones, presumably by counteracting N-cadherin or LAR function. We are currently cloning ovs in the hope that its molecular identity will provide insight into the regulatory mechanisms of R7 target selection. [unreadable] [unreadable] As layer-specific connectivity, the retinotopic map is formed in two separate stages. In the larval stage, a coarse topographic map is established by two partially redundant mechanisms: one is mediated by the secreted protein DWnt4 and its receptor Dfrizzled2, while the other is by afferent-afferent and afferent-target interactions mediated by various adhesion molecules. In the past few years, we studied how the R7 retinotopic map is maintained and further refined at the pupal stages. In a genetic screen based wavelength-selection behavior, we identified pex (premature extension), which affects the refinement of R7 retinotopic map. Pex is a novel allele of baboon, which encodes a type I activin receptor. Baboon mutant R7 axons target to the appropriate retinotopic medulla column in the correct layer, but they extend collaterals to innervate neighboring medulla columns, indicating that the map formation, but not the layer-specific targeting, is disrupted.[unreadable] [unreadable] We found that the canonical activin signaling components are conserved in R7s. Mutations disrupting dSmad2, which encodes the transcription factor downstream of baboon, result in baboon-like R7 phenotypes. In a collaboration with Dr. Tory Herman, we found that importi-alpha3 mutants exhibit R7 retinotopic map defects essentially identical to those in baboon/dSmad2 mutants. Importin-alpha3 and dSmad2 are present in the R7 growth cones and they form physical complexes. Removing importin-alpha3 in R7s disrupts normal nuclear accumulation of dSmad2, suggesting that importin-alpha3 is required for nuclear import of dSmad2. While activin is expressed in both R7s and subsets of medulla neurons, disrupting activin function only in R7s using RNAi or dominant negative construct results in baboon-like R7 phenotypes. Together, these data suggest that auto/paracrine activin refines R7 retinotopic map through baboon and dSmad2/importin-?3 complexes.